Re: icu4c api proposal: clone() for Replaceable and UnicodeString

[Markus S. <mar...@jt...>]

Another reply to myself...

I am struggling to define the semantics for the C API UReplaceable close function; for example, a 
user should probably never free() the "object" even if the close function pointer is NULL...

Anyway, the C UReplaceable only exists to facilitate the C Transliterator API (utrans.h), and a 
transliteration/transform operation will never clone its replaceable. The UReplaceable is likely 
statically or stack-allocated. So, the clone() seems pointless for this C API. Just extra work right 
now...

Therefore, I retract my proposals for new UReplaceable clone/close functions for now and will 
implement clone() only for the C++ Replaceable class tree.

markus

Markus Scherer wrote:
> Markus Scherer wrote:
> 
>> UReplaceableCallbacks will get:
>>
>>   UReplaceable * (*clone)(const UReplaceable* rep);
> 
> It turns out that for this to work in the C API, one has to be able to 
> get rid of the clone. However, the UReplaceableCallbacks struct does not 
> currently contain a "destructor" function pointer. Therefore, I also 
> propose:
> 
>   void (*close)(UReplaceable* rep);
> 
> markus